Bearing monitoring is not a single measurement. It is a system: the right sensor at the right location, measured consistently at the same load, compared against a baseline from when the bearing was new, with alarm levels that trigger action before failure, not after.
Step 1: Identify Bearings to Monitor and Their Priority
Not every bearing warrants continuous or even periodic monitoring. Start by identifying where a failure actually matters.
Prioritize bearings whose failure causes production loss or safety risk; bearings on assets with no installed redundancy; bearings in locations where replacement requires significant labor or machine downtime; and bearings on assets with a documented history of bearing failures.
For each prioritized bearing, document the following before any sensor goes on the machine:
- Bearing designation and location (drive end, non-drive end, pump bearing)
- Shaft speed in RPM at typical operating conditions
- Radial load direction
- Accessibility for sensor mounting
This documentation becomes the foundation for fault frequency calculation, sensor placement decisions, and measurement scheduling. Skipping it means the rest of the program is built on incomplete information.
Step 2: Sensor Selection and Mounting
For most industrial rotating machinery above 60 RPM, a 100 mV/g ICP accelerometer is the standard choice. Stud-mount wherever possible; adhesive mounting is acceptable for non-critical assets or locations where drilling is not permitted, but adhesive reduces the usable frequency range and should be documented as a constraint.
For slow-speed equipment below 60 RPM, standard accelerometers lose sensitivity at the low frequencies where bearing defects appear. Use a high-sensitivity accelerometer (500 to 1000 mV/g) or an ultrasound sensor, which detects the stress wave emissions from early-stage defects at low speeds where vibration amplitude is still very small.
Mounting Location
Place the sensor as close to the bearing as physically possible, on the load zone side of the bearing housing. For a motor with a horizontal shaft carrying gravity load, the load zone is at the bottom of the bearing; the sensor should be at the 6 o'clock position on the bearing housing.
Take measurements in three directions at each bearing point:
- Radial horizontal (perpendicular to shaft, horizontal plane)
- Radial vertical (perpendicular to shaft, vertical plane)
- Axial (along the shaft axis)
Axial readings are often skipped in abbreviated programs, but they are the primary indicator of thrust bearing defects and misalignment-induced axial loading. Omitting them creates a blind spot.
Step 3: Calculate Expected Fault Frequencies
Before collecting the first measurement, calculate the four bearing defect frequencies using the manufacturer's geometry data for each bearing in the program:
| Frequency | What It Detects | Input Data Required |
|---|---|---|
| BPFO (Ball Pass Frequency, Outer Race) | Outer race defect | Number of balls, contact angle, ball diameter, pitch diameter, shaft speed |
| BPFI (Ball Pass Frequency, Inner Race) | Inner race defect | Same as BPFO |
| BSF (Ball Spin Frequency) | Rolling element defect | Ball diameter, pitch diameter, contact angle, shaft speed |
| FTF (Fundamental Train Frequency) | Cage defect | Same as BPFI |
Mark these frequencies as the targets to watch in the vibration spectrum. Many monitoring systems accept the bearing designation number and compute fault frequencies automatically from a built-in database. If using a portable analyzer with manual entry, calculate these before the first site visit and enter them into the measurement point configuration.
Without fault frequencies pre-loaded, a spectrum shows a waveform with no context. The analyst has to recognize elevated energy at a specific frequency and know what it means. Pre-loaded fault frequencies make the connection explicit and reduce the skill required to interpret each reading.
Step 4: Establish a Baseline
The baseline defines what normal looks like for that specific bearing at that specific operating condition. Every future reading is compared against it. A poor baseline produces false alarms and missed faults in roughly equal measure.
Take the baseline measurement under these conditions:
- Known representative load: ideally full load or the most common operating load. Partial load measurements produce lower amplitudes that will generate false alarms when the machine runs at full load.
- Stable machine temperature: after at least 30 minutes of continuous operation. Cold starts produce transient vibration signatures that do not represent steady-state behavior.
- Sensor firmly mounted at the defined measurement point: document the exact location with a photo or sketch so every subsequent measurement is taken at the same position.
Record the load condition and ambient temperature alongside the baseline data. Both affect vibration amplitude and will be needed to interpret readings taken at different conditions later.
The baseline must be taken on a bearing in known good condition: immediately after a confirmed successful installation of a new bearing, or after a rebuild with documented bearing inspection results. A baseline on a bearing of unknown condition measures an unknown state and provides no useful reference.
Step 5: Set Alarm Levels
Two levels are sufficient for most programs: advisory and danger. Adding more levels adds administrative overhead without proportional diagnostic value.
Frequency-Specific Alarms (at BPFO, BPFI, BSF, FTF)
- Advisory: amplitude at fault frequency reaches 150% of baseline
- Danger: amplitude at fault frequency reaches 250% of baseline
These are starting points. Adjust based on experience with each machine type, the rate at which faults have historically progressed on similar assets, and the consequences of failure if the machine runs past the danger threshold.
Overall Vibration Alarms
Use ISO 10816 zone boundaries as a cross-check on overall vibration level. Zone C boundary triggers advisory; Zone D boundary triggers danger. These are broad indicators that confirm something is changing, even before a specific fault frequency emerges clearly.
Temperature Alarms
- Advisory: operating temperature rises 10 C above the established baseline
- Danger: operating temperature rises 20 C above baseline
Temperature alarms catch lubrication failures and advanced mechanical defects that vibration may not flag early. They are a low-cost complement to vibration monitoring, not a replacement.
Step 6: Measurement Schedule
The right measurement frequency depends on how fast faults progress on each asset class. There are two delivery models:
Continuous Monitoring
Permanently mounted sensors transmit data in real time. Appropriate for critical assets where a failure between inspections is unacceptable: process-critical pumps, compressors, main drive motors, and any asset where an unplanned stop costs more than the monitoring infrastructure.
Route-Based Periodic Monitoring
A technician visits each asset on a schedule with a portable analyzer. Starting intervals:
| Asset Class | Starting Interval |
|---|---|
| Critical (failure causes immediate production loss) | Monthly |
| Semi-critical (failure causes partial or delayed impact) | Quarterly |
| Non-critical (redundancy available or low consequence) | Semi-annually |
These are starting points. Adjust intervals based on observed alarm frequency and how quickly fault indicators progress on each specific machine. An asset that consistently shows rapid progression from advisory to danger should move to a shorter interval. One that runs stable for quarters with no trending can be extended.
Step 7: Alarm Response Protocol
Define the response before the alarm fires. Teams that decide what to do after an alarm appears lose time to discussion and often under-react or over-react depending on who is on shift. The protocol should be written, accessible, and referenced in work orders.
Advisory Alarm
- Increase measurement frequency: for route-based programs, move to weekly
- Investigate potential root cause: review recent operational history, process upsets, overloads, recent starts
- Check lubrication: verify type, quantity, and interval are correct for the application
- Document findings and open a work order to track the developing condition
Danger Alarm
- Schedule bearing replacement at the next available maintenance window
- Do not restart without an engineering assessment if the machine is already shut down
- Verify with an additional measurement to confirm the reading is not an artifact of sensor looseness or transient process condition
Immediate Assessment Threshold
- If vibration reaches ISO 10816 Zone D, assess whether immediate shutdown is warranted based on the failure consequence and production context
- Any bearing running hot enough to discolor the housing or produce a burning smell warrants immediate shutdown regardless of measured vibration level
The goal of the protocol is to remove ambiguity from a situation where people are under time pressure. Defined thresholds with defined actions eliminate the "how bad is it really?" conversation that delays response.
Step 8: Documentation and Trending
A bearing monitoring program produces value twice: once when it catches a fault in progress, and again when the historical record informs decisions about future assets.
Maintain a record for each bearing point that includes:
- All measurement dates and readings, with the operating conditions at time of measurement
- Alarm events, their severity, the investigation findings, and the resolution
- Bearing replacement records with the failure mode observed on the removed bearing
- Lubrication events: type, quantity, date
The record establishes the fault progression rate for that asset class. If a specific pump model consistently moves from advisory to danger over eight weeks, the measurement interval for similar pumps should be four to six weeks, not monthly. If another asset class sits at advisory for six months without progressing, the interval can be extended.
Over time, the program becomes self-calibrating. The data drives interval decisions rather than general industry guidelines that may not reflect the actual operating environment, load cycles, and maintenance practices at a specific facility.
The failure mode observed on the removed bearing closes the loop: it confirms whether the fault frequency that triggered the alarm was correctly identified, and it trains analysts to recognize the same signature earlier on the next asset.
Automate Steps 3 Through 7
Tractian's sensors implement the core steps of this program automatically: fault frequencies computed from bearing data, baselines established from first measurements, and alarms triggered when amplitude trends exceed defined thresholds.
See Mechanical Monitoring